canva-creator
Takes an approved content brief and executes a campaign end-to-end: builds the posting calendar, generates Canva designs for social posts, drafts caption and email copy, and stages social sends in HubSpot.
Install / Use
npx skills add anthropics/knowledge-work-plugins --skill canva-creatorInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
CommunicationSupported Platforms
Our assessment of canva-creator
canva-creator scores 97/100 on our quality scale, 12th of 170 Communication skills we index (top 8%).
Its SKILL.md is 19 KB long, well organised into 13 sections with 3 code examples: a thorough specification that gives an agent plenty to work with.
With 25,526 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated yesterday, so canva-creator is actively maintained.
- It is released under the Apache-2.0 license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands.
Automated pattern scan on 2026-09-26. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
canva-creator compared with similar skills
All 4 of these similar skills score higher than canva-creator; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| canva-creator (this skill)by anthropics | 97 | 25.5k | 1d ago | SKILL.md |
| algorithmic-artby anthropics | 100 | 177.9k | 3d ago | SKILL.md |
| pptxby anthropics | 100 | 177.9k | 3d ago | SKILL.md |
| designby nextlevelbuilder | 100 | 130.2k | 4d ago | SKILL.md |
| ui-ux-pro-maxby nextlevelbuilder | 100 | 130.2k | 4d ago | SKILL.md |
Frequently asked questions
- How do I install canva-creator?
- Run
npx skills add anthropics/knowledge-work-plugins --skill canva-creator. The install tabs above show the steps for each supported agent. - Which AI agents does canva-creator work with?
- It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is canva-creator safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. It is Apache-2.0-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is canva-creator still maintained?
- The repository was last updated yesterday, so canva-creator is actively maintained.
Skill content
View source on GitHubname: canva-creator description: > Takes an approved content brief and executes a campaign end-to-end: builds the posting calendar, generates Canva designs for social posts, drafts caption and email copy, and stages social sends in HubSpot. Canva is used for social posts only (Instagram, Facebook, X, LinkedIn) — email content is drafted as plain text and surfaced inline for the owner to send from their own tool. Every step requires explicit owner approval. This is the one-shot campaign executor: a finished brief goes in, a complete campaign comes out. Use when the owner hands off an approved brief or points at one — "run this brief," "execute the campaign," "turn this brief into posts," "build the campaign from the plan we approved." For day-to-day posting with no brief — "I need to post more," "what should we post" — use social-content-engine instead; it owns the standing calendar. allowed-tools: Read, WebFetch
Canva Creator
Scope
This skill handles a campaign in five sequential stages, each gated by owner approval:
brief → calendar → asset inventory → Canva designs → copy → HubSpot staging
| Path | Channels | What this skill produces | |------|----------|--------------------------| | Canva (social) | Instagram, Facebook, X/Twitter, LinkedIn | Canva design + caption + scheduled HubSpot post | | Text-only | Email (newsletter, marketing, drip) | Subject + preheader + body, surfaced inline for the owner to send |
Canva is not used for email rows under any circumstance — no templates,
no autofill, no design copies, no asset uploads, no exports. Email
graphics are out of scope for this skill because email-template
autofill produces placeholder graphics when image slots exceed available
photos, and variation thumbnails fail to render in chat previews. If the
owner asks for a Canva email design, see reference/gotchas.md for the
redirect language.
Pre-flight
Before Stage 1, confirm:
-
Brief. The user has referenced or pasted an approved brief. If not: "I'll need the content brief before I can build the campaign. Do you have one from the content-strategy skill, or would you like to write one now?"
-
Canva tier. Pro/Teams require manual template selection from the user's library (no autofill API). Enterprise can autofill from brand templates.
-
HubSpot tier. Social staging requires Marketing Hub Professional. Starter or Free → skip Stage 5 and export a CSV instead (see reference/hubspot-staging.md).
-
Brand assets. If a storefront is connected (Shopify or Square — peers,
../../shared/connector-neutrality.md), read the product list first: titles, image URLs, and prices for every product the brief names. Read only — nothing is uploaded to Canva until Stage 2 step 5, and then only the images the approved calendar's slots need. That list is the starting photo inventory; ask only about products the store has no image for. The pull rules — which fields, the price-claim rule — are in ../social-content-engine/reference/shopify-assets.md (written for Shopify; Square's catalog images follow the same steps from the tool inventory). With no storefront, confirm the path to product photos on disk or that the brand kit is live in Canva. -
Generation budget. Estimate the campaign's Canva volume and surface it before Stage 1 begins. Default is 3 candidates per Canva-bound row; each design costs ~5 API calls (autofill + export + polling).
Generation budget for this campaign: Canva (social) rows: 8 Candidates per row: 3 (default — say "single candidate" to use 1) Total designs: 24 API calls (approx): ~120 (autofill + export + polling) Canva limit: 100 requests/minute. This will take ~2-3 minutes of generation, well within your tier limits. Proceed?If the projected total designs exceeds 30, recommend single-candidate mode upfront — large campaigns run out of headroom fast. The owner can override the default to 1, 2, or 3 candidates per row before Stage 1 starts. Lock the chosen value for the entire session.
Workflow
Stage 1 — Posting calendar
Pull from the brief: content themes, channels, cadence, hard dates (launches, sales, holidays).
Build a calendar table with a Path column that routes every row to either
Canva or text-only drafting:
| Date | Channel | Path | Theme | Asset type | Caption/Subject angle | |------|---------|------|-------|------------|-----------------------| | Jun 2 | Instagram feed | Canva (social) | Linen launch | Square post | "finally, a dress…" | | Jun 5 | Email | Text-only | Linen launch | Email body | "Linen that actually breathes" |
Tag every email-channel row as Text-only before presenting. Cap at 30
days unless the brief specifies otherwise. Flag scheduling conflicts (two
posts same day for the same product) up front.
Checkpoint 1. Present the calendar. Ask: "Does this match the plan? Any dates to shift, channels to add, or themes to swap?" Iterate until approved, then restate the split out loud — "N rows go through Canva, M rows go through text-only drafting" — before moving on. Catching a miscategorization here is free; catching it after generating designs isn't.
Stage 2 — Asset inventory (Canva rows only)
Email rows skip this stage entirely. For each Canva (social) row, build
a manifest of what the template needs and what's already available.
-
Enumerate every image slot by name. Square Instagram posts usually have 1-2 image slots; carousels and product grids can have 5+. List them individually (
Header_Image,Product1_Image,Product2_Image, …) — never roll them up as "product images."- Enterprise: read field names from
dataset[].labelon the brand template (GET /v1/brand-templates/{id}). - Pro/Teams: count every distinct image rectangle in the template.
- Enterprise: read field names from
-
Inventory available assets. Text content from the brief (product names, offer copy, taglines, pricing), product image URLs read from the connected storefront at pre-flight (uploaded in step 5, only for the slots the gap table needs, and recorded by
asset.idthen), product photos already uploaded to Canva (GET /v1/assets) or on the owner's disk, brand kit colors and fonts (Enterprise). A price in a caption comes from the storefront's returned value or the brief, never from memory. -
Build the slot-by-slot gap table. One row per slot per design — not per design.
| Date | Slot name | Slot kind | Available asset | Status | |------|-----------|-----------|-----------------|--------| | Jun 2 | Hero_Image | image | bloom_summer.jpg → asset_id pending | upload | | Jun 2 | Headline | text | "Summer linen, finally" | ready | | Jun 9 | Product1_Image | image | — | MISSING |
-
Resolve slot/asset mismatches with the owner. If the template has more image slots than the brief provides photos, pause and ask:
The "Summer Carousel" template has 5 image slots. The brief gave me 1 photo (bloom_summer.jpg). How should I fill the other 4? 1. Reuse the same photo across all 5 slots 2. You send me 4 more photos (file paths) 3. Pick a simpler template with fewer slotsNo generation calls until the owner picks. Generating with empty slots produces designs full of Canva's default landscape placeholders.
-
Upload missing photos and capture verified asset IDs. Upload via
POST /v1/asset-uploads, then pollGET /v1/asset-uploads/{job_id}untilstatus == "success". Recordasset.idfrom the response — this is the only value that works in an autofill image field. Passing an empty string, a URL, a file path, or a stale ID silently renders Canva's stock landscape graphic instead of the photo. -
Confirm the manifest. Show the owner the completed slot-by-slot table with every slot resolved and every image
asset.idconfirmed. This is the last stop before Canva API calls.
Stage 3 — Canva design generation
Before any Canva API call, re-read the calendar and drop any row whose
Path is not Canva (social). Email rows do not pass through this
stage.
Generate designs one calendar row at a time, with 3 candidates per row (or the value chosen at pre-flight). Each row follows the same loop: generate candidates → verify → export → visually check → retry failures → present → wait for owner pick → next row. Pause 30 seconds between rows. This caps the burst at 3 generations + 3 exports per ~30s — well under Canva's 100 req/min rate limit. Do not parallelize multiple rows; one row at a time is the protection that keeps the owner from hitting quota mid-campaign.
Polling cadence. Poll job status every 3-5 seconds, not faster. Tighter intervals burn quota without speeding up completion.
Preview URLs — only one type is safe to embed. Autofill responses
return design.canva.ai thumbnails that expire within minutes; embedding
them as markdown images produces broken "Show Image" placeholders.
Permanent export URLs (export-download.canva.com or the export-design
MCP tool) do not expire. Native Cowork carousels render the autofill
result directly using the connector's authenticated session — let them
render on their own, don't re-embed.
Row loop
-
Resolve template. (Once per session — same template across rows unless the calendar mixes asset types.)
- Enterprise:
GET /v1/brand-templatesfiltered by asset type. - Pro/Teams:
GET /v1/designs?ownership=any&query={template name}, surface top 3 to the owner, confirm one before generating.
- Enterprise:
-
Generate the row's candidates in parallel. Fire the row's 3 candidates simultaneously (or N from pre-flight).
- Enterprise:
POST /v1/autofillsper candidate with the template ID and field values. Poll all jobs concurrently. - Pro/Teams:
POST /v1/designsto create copies. Describe the text and image edits the owner applies in Canva; collect design IDs back.
- Enterprise:
-
Verify job status. For each candidate, confirm
GET /autofills/{job_id}returnedstatus == "success"andresult.design.idis present. Handle errors per-design:-
JOB_FAILED→ readjob.error.message, fix the field values or asset IDs, retry once. -
RATE_LIMIT_EXCEEDED(first hit this session) → wait 60s, retry that one candidate once. This handles transient spikes. -
RATE_LIMIT_EXCEEDED(second hit this session) or anyquota_exceeded/ daily-cap error → stop generation immediately. Do not retry. Surface progress and ask:Canva is rate-limiting the campaign. Status so far: ✓ Generated: Posts 1-4 (12 designs) ⏸ Remaining: Posts 5-8 (12 designs not yet generated) How should I proceed? 1. Switch to 1 candidate per remaining row (4 designs total) — finishes now 2. Pause campaign — resume in 60 minutes when quota refills 3. Stop generation — work with what we have, move to captionsWait for the owner's choice. Do not loop on retry.
-
-
Export each successful candidate to a permanent PNG. Fire the row's exports in parallel.
- REST:
POST /v1/exportswithformat.type: "png", pollGET /v1/exports/{job_id}until success, captureurls[0]. - Canva MCP:
export-designwith the design ID.
These permanent URLs are what get embedded in previews and attached to the HubSpot post later. The autofill response thumbnail is never used downstream.
- REST:
-
Visually verify each export. Look at the image and reject any of these — they all indicate an unfilled slot or wrong asset:
- Generic landscape with clouds and green hills (Canva's default placeholder)
- Solid gray rectangles where a photo should be
Truncated for display — read the full file on GitHub.
Related Skills
algorithmic-art
177.9kCreating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems.
pptx
177.9kUse this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an em…
design
130.2kComprehensive design skill: brand identity, design tokens, UI styling, logo generation (55 styles, Gemini, Atlas Cloud, or MuAPI AI), corporate identity program (50 deliverables, CIP mockups), HTML presentations (Chart.js), banner design (22 styles, social/ads/web/print), icon design (15 styles, SVG…
ui-ux-pro-max
130.2kUI/UX design intelligence for web, mobile, and desktop. This skill should be used when designing, building, reviewing, or fixing interfaces, including pages, components, design systems, accessibility, interaction, responsive layout, typography, color, charts, and stack-specific UI implementation.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
